Temperate
Legal

Subprocessors

Last updated: 8 November 2026

Draft. Our legal adviser is still reviewing this text, so it is not final and binds no one yet. Parts in [square brackets] are still open. Questions go to [email protected].

These companies process the data of our customers for us. Each one is bound by a written agreement to the same duties we have under ourdata processing agreement. None of them uses your data to train AI models.

The list

Amazon Web Services
What it does: Hosting, the database, file storage, the virus scan of uploads, the sandbox that runs code, and AI models through Amazon Bedrock.
Data: All content of your organization.
Where: Frankfurt, Germany, and for AI models other AWS regions in the EU. Files reach your browser through Amazon CloudFront, which delivers them from locations in Europe and North America, under the EU-US Data Privacy Framework.
OpenAI
What it does: AI models, the search index of your knowledge, and images.
Data: The messages, files and passages a task needs.
Where: EU. Every call goes to the EU endpoint.
Jina AI
What it does: Web search, and ranking the passages that answer a question best.
Data: Search queries, questions and passages of your documents.
Where: EU.
Linkup
What it does: Fetching the web pages an agent opens.
Data: The addresses of those pages.
Where: EU.
Brevo
What it does: The emails Temperate sends, such as invitations and notices.
Data: Email addresses, names, and the text of the email.
Where: EU.
Sentry
What it does: Error reports, so we can fix what breaks.
Data: Technical details of an error, which can hold content.
Where: EU.

Systems you connect

When you connect a system, such as Google, Microsoft 365 or Exact, or add your own MCP server, Temperate sends it what an agent needs to act there. You choose these systems, and the terms of their provider apply.

Changes

Before a new subprocessor starts, we tell the owners and admins of each organization by email, and update this page. You can object to the change. Questions go to [email protected].